Why you need to upgrade to a knowledge management system

AI automation is one of the biggest ever breakthroughs in business. It has the potential to enable anyone to run a global business, and that’s exactly what we’re building a superhuman business partner to do. 

If you’ve read anything about agents, it’s likely focussed on what assistants and agents can and will be able to do. This makes sense, it’s the executional aspect which most people can understand first. It’s why you’ll see simple demos of an agent that can book you tickets, or an agent that can automatically sort your meetings. These are executional examples that are tangible because we’ve all done them ourselves. But how will those agents know what decisions to make or what is right for our business?

Understanding the difference between Assistants and Agents

When working with AI it’s firstly important to understand the difference between assistants and agents. Assistants can reference knowledge, give guidance or take actions on your behalf when asked. Agents, on the other hand, are set specific roles and capabilities and are able to execute tasks automatically on their own when they deem it necessary. That’s a subtle but important difference and there are many people out there confusing the two, often somewhat intentionally to exaggerate their product’s capabilities. If it's autonomous, it's an agent. If it's not, it's probably an assistant.

Trust me, the world of agents is coming, we’re building for it and if you’re in business you need to prepare for it. But to do so you have to go beyond the executional concepts and ask yourself an important question – what foundation does an agent need to be successful across my business, how will we direct them, and how will they know what we need? 

Understanding knowledge management and knowledge systems

If AI is executing tasks automatically for us, the answer to these questions lies in how we will interface with it. The most important aspect of human and AI interaction in the future will be knowledge management and distribution. It may not sound sexy, but it will become the sole layer where we will interact directly with AI. If you’ve played with AI it’s likely you have come across the concept of knowledge, put simply it is information that the AI can reference that is relevant to you or your business. This is the key layer to understand if you want to prepare your business for automation and the future. Most people are currently operating at what is best described as rudimentary knowledge management: dumping links & files into a repository and having an LLM go and attempt to find what is relevant to a request. 

This can feel empowering at first, it’s most people’s big wow moment with AI. Wow, it came back with something that’s relevant! But there are fundamental flaws in this approach which quickly unravel. Firstly, those links, files and documents will inevitably contain conflicting variations on pieces of relevant information. For instance, something might be described differently, have different pricing, contain mistakes or vague language. These variations will expand exponentially based on business size, and I’ve previously written about how this will inevitably make full-scale automation completely impossible in enterprise businesses. Secondly, with this approach you have little to no chance of having an agent be able to accurately reference or update knowledge for you without returning errors.

For a good analogy, imagine you are walking down the street and you meet two people walking together. You ask them which way to the pub, one says left and one says right. Which way do you go? There is only one way to a restaurant, but you now cannot discern which piece of information is right. Now imagine you meet a crowd of 100 people and you ask them which way and all at the same time they shout 100 different directions, how do you determine the right way? If you're dumping information as knowledge – this is what you're essentially asking AI to do.

There’s a reason Google maps is built on a single source of truth.

Most organizations are stuck at this extremely basic level of AI implementation, which is really just hopeful information retrieval. It's the lowest level of automation and a good tell as to whether an organization is doing this is they have created a 'knowledge base'. If this is you, then it's time to step it up a level - because there is a better way.

Why every business needs a knowledge system

To successfully automate a company with AI you need to understand knowledge systems and how they differ from a basic knowledge base. Firstly to be accurate, AI needs a single source of truth to operate from. This is not a single repository of dumped files and information, it’s a single source of truth built on individual facts of a business. Each fact in this single source of truth is independent and just like in the real world there can be no two versions of the same fact.

Secondly, many of those facts impact other facts, as a result you need to have a system which connects the elements of knowledge across the business. A single change to knowledge in a knowledge system should immediately ripple across it, changing output and responses for assistants, agents and any team member creating content with AI. An agent, for example, is then always operating with the latest piece of knowledge and can therefore reason what to do accurately based on its role and tasks.

If you want to automate your business it is important to reshape your view of what your business actually is. Ultimately your business is just knowledge built out of individual facts and your role is to determine which facts you want to use to direct AI. 

As a result of this, the most important aspect of the automated tech stack in business right now is the knowledge system that sits at the center of your ecosystem and is controllable by both you and the AI assistants or agents that you deploy from it. If you don’t have one yet, you will need one. This is exactly why we built NOAN as an AI native operating system for business with knowledge at the center of the experience, it’s the only way that true automation can occur.

Our vision is a world where you can control your entire company via semantic direction, ultimately voice. It will rely on humans collaborating with AI via knowledge control and knowledge management. Come try it out and join us on the journey.